feat(H): H-EXT/H-ENT/H-AUTO/H-CONF — LLM knowledge extraction (entities, relationships, auto-create in GraphRAG, confidence scoring with review queue), 31 tests passing

tests/test_phase_h_wiki.py

@@ -275,3 +275,103 @@ class TestEvidenceReferences:
 
         refs = build_evidence_references([])
         assert refs == []
+
+
+# ─── H-EXT/H-ENT/H-AUTO/H-CONF: Knowledge Extraction ─────────────────────────
+
+
+class TestKnowledgeExtraction:
+    """Test the knowledge extraction module (H-EXT, H-ENT, H-AUTO, H-CONF)."""
+
+    def test_extracted_entity_dataclass(self):
+        """ExtractedEntity dataclass works correctly."""
+        from app.ai.knowledge_extraction import ExtractedEntity
+
+        entity = ExtractedEntity(name="John Doe", entity_type="person", confidence=0.9)
+        assert entity.name == "John Doe"
+        assert entity.entity_type == "person"
+        assert entity.confidence == 0.9
+        assert entity.mentions == []
+        assert entity.metadata == {}
+
+    def test_extracted_relationship_dataclass(self):
+        """ExtractedRelationship dataclass works correctly."""
+        from app.ai.knowledge_extraction import ExtractedRelationship
+
+        rel = ExtractedRelationship(
+            source_entity="John Doe", source_type="person",
+            target_entity="Acme Corp", target_type="company",
+            relationship_type="works_for", confidence=0.85,
+            evidence="John works at Acme",
+        )
+        assert rel.source_entity == "John Doe"
+        assert rel.relationship_type == "works_for"
+        assert rel.confidence == 0.85
+        assert rel.evidence == "John works at Acme"
+
+    def test_extraction_result_dataclass(self):
+        """ExtractionResult dataclass works correctly."""
+        from app.ai.knowledge_extraction import ExtractionResult
+
+        result = ExtractionResult(source_type="wiki", source_id="abc", tenant_id="123")
+        assert result.entities == []
+        assert result.relationships == []
+        assert result.overall_confidence == 0.0
+
+    def test_is_low_confidence(self):
+        """is_low_confidence correctly identifies low confidence scores."""
+        from app.ai.knowledge_extraction import is_low_confidence, LOW_CONFIDENCE_THRESHOLD
+
+        assert is_low_confidence(0.3) is True
+        assert is_low_confidence(0.5) is True
+        assert is_low_confidence(0.59) is True
+        assert is_low_confidence(0.6) is False
+        assert is_low_confidence(0.9) is False
+        assert LOW_CONFIDENCE_THRESHOLD == 0.6
+
+    def test_filter_high_confidence(self):
+        """filter_high_confidence splits relationships correctly."""
+        from app.ai.knowledge_extraction import filter_high_confidence, ExtractedRelationship
+
+        rels = [
+            ExtractedRelationship("A", "person", "B", "company", "works_for", 0.9),
+            ExtractedRelationship("C", "person", "D", "company", "related_to", 0.3),
+            ExtractedRelationship("E", "person", "F", "company", "knows", 0.7),
+            ExtractedRelationship("G", "person", "H", "company", "met", 0.5),
+        ]
+        high, low = filter_high_confidence(rels)
+        assert len(high) == 2  # 0.9 and 0.7
+        assert len(low) == 2   # 0.3 and 0.5
+        assert high[0].confidence == 0.9
+        assert high[1].confidence == 0.7
+
+    def test_filter_high_confidence_custom_threshold(self):
+        """filter_high_confidence respects custom threshold."""
+        from app.ai.knowledge_extraction import filter_high_confidence, ExtractedRelationship
+
+        rels = [
+            ExtractedRelationship("A", "person", "B", "company", "works_for", 0.8),
+            ExtractedRelationship("C", "person", "D", "company", "related_to", 0.7),
+        ]
+        high, low = filter_high_confidence(rels, threshold=0.75)
+        assert len(high) == 1  # 0.8
+        assert len(low) == 1   # 0.7
+
+    @pytest.mark.asyncio
+    async def test_extract_knowledge_empty_text(self):
+        """extract_knowledge returns empty result for empty text."""
+        from app.ai.knowledge_extraction import extract_knowledge
+
+        result = await extract_knowledge("", uuid.uuid4())
+        assert result.entities == []
+        assert result.relationships == []
+        assert result.overall_confidence == 0.0
+
+    @pytest.mark.asyncio
+    async def test_extract_knowledge_short_text(self):
+        """extract_knowledge returns empty result for very short text."""
+        from app.ai.knowledge_extraction import extract_knowledge
+
+        result = await extract_knowledge("Hi", uuid.uuid4())
+        assert result.entities == []
+        assert result.relationships == []